Joint CC and Bimax: A Biclustering Method for Single-Cell RNA-Seq Data Analysis

One of the important aims of analyzing single-cell RNA sequencing (scRNA-seq) data is to discovery new cell subtypes by clustering. For the scRNA-seq data, it is obvious that lots of genes have similar behavior under the different conditions (cells). Traditional clustering algorithms could not obtain high-quality cluster on scRNA-seq data. However, the biclustering algorithm has begun a more powerful data mining tool, which can cluster genes and conditions (cells) simultaneously. In this paper, we propose a novel biclustering algorithm named JCB: Joint CC and BIMAX. The algorithm is based on the two classic biclustering algorithms: Cheng and Church’s algorithm (CC) and Binary Inclusion-Maximal biclustering algorithm (Bimax). The main idea of the JCB method is that it joints the “mean squared residual (MSR)” proposed by CC with the model of BIMAX. It gets the biclusters by iterating on rows and columns of the data matrix with the “MSR”, and it also benefits the advantage of simply model from Bimax. We evaluate the proposed method by carrying out extensive experiments on three scRNA-seq datasets. The JCB method is used to compete with six other bi-clustering algorithms. The experimental results show that the proposed method outperforms the others.